Add reachability checks and develop/main container pipeline (inactive)

- preflight --live: Ollama/Qdrant reachable from the container, required
  Ollama models present, 768 dimensions; clear abort instead of traceback.
  Static check rejects unset or container-local service URLs.
- deploy/smoke.py: non-writing post-start check via published ports.
- deploy/deploy.sh: test (no network), build, preflight, up --wait, smoke;
  prod rolls back to the last good image tag.
- Workflow templates deploy-dev.yml (push develop) and deploy-prod.yml
  (push main) following the Kansho pattern; not under .gitea/workflows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
LarsandClaude Opus 5.5 committed 2026-10-28 16:01:37 +01:00
1 parent 7a225ba2b6
commit 8f9f83aa79
14 files changed
+728 -90

No files matched your search

+33
View File
@@ -0,0 +1,33 @@
# VORLAGE, NICHT AKTIV: liegt absichtlich nicht unter .gitea/workflows/.
# Aktivierung erst bei der Umschaltung (Gate G5, eigene Freigabe), und zwar im
# selben Commit, der .gitea/workflows/deploy.yml (systemd) entfernt:
# git rm .gitea/workflows/deploy.yml
# git mv deploy/gitea-workflows/deploy-prod.yml .gitea/workflows/deploy-prod.yml
# Sonst deployen bei jedem Push auf main beide Wege.
#
# Prod nur über Merge-Commit develop -> main (git merge --no-ff develop), wie bei Kansho.
# Ablauf, Abbruchbedingungen und Rollback: deploy/deploy.sh.
name: Deploy mindnet PROD (Container)
on:
push:
branches: [main]
workflow_dispatch:
concurrency:
# Gleiche Gruppe wie der systemd-Workflow: nie zwei Prod-Deploys gleichzeitig.
group: deploy-mindnet
cancel-in-progress: false
jobs:
deploy:
runs-on: linux_host
env:
MINDNET_ENV_FILE: /home/llmadmin/docker/mindnet/prod.env
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Test, Build, Preflight, Start, Smoke (Rollback bei Fehler)
run: bash deploy/deploy.sh prod